2010 - Nvidia Fermi: The First GPU to Power World's Highest-Performing Supercomputer
General purpose computing on graphics processing unit (GPGPU) is the idea that is almost ten years old. Thanks to massively parallel architecture, GPUs are great for parallel computing. This year Nvidia supplied compute cards to power the world's most powerful supercomputer.
Back in 2006 - 2007 both ATI, graphics business unit of Advanced Micro Devices, and Nvidia introduced their specially-designed compute cards aimed at high-performance computing (HPC) markets. Besides, Nvidia spent a lot of resources on creation of its CUDA platform for GPU computing and also integrated a number of compute-specific logic into its Fermi-series graphics processors (for example, the Fermi architecture is tailored to deliver maximum double-precision floating point computer performance; SIMD processors of Fermi can both read and write from/to the unified L2 cache, something that is needed for compute and something that is less needed for graphics; etc). The approach of ATI-AMD was much more pragmatic: instead of creating a proprietary GPGPU platform the firm decided to rely on industrial standards like DirectCompute or OpenCL; instead of building-in excessive compute-specific functionality into Cypress (Radeon HD 5000), the company chose to integrate only the most important ones to keep the size of the chip smaller and wait for new process technologies to integrate new features.
As the more GPGPU-concentrated player, Nvidia was also the first to receive fruits from its efforts. In 2010 three supercomputers out of top five supercomputers in the world, including the #1 Tiahne-1A with 2.56petaFLOPS performance, were powered by Nvidia Tesla 2000-series compute cards. Moreover, the same cards were used in many other HPC systems.
The market of GPU-based accelerators for HPCs is not large; financially Nvidia still earns the majority of its revenue by selling consumer GeForce-branded graphics cards and Quadro-branded professional graphics accelerators. But what is important is that there is general belief that supercomputers capable of performing at least a quintillion double-precision floating point operations per second (1018 FLOPS of ExaFLOPS) should be heterogeneous, e.g. employ both highly-parallel as well as high-performance serial processors. The Tianhe-1A is among the first heterogeneous supercomputers and the fact that is already the world's most powerful supercomputer may not be a turning point, but it is a clear signal for everyone in the HPC industry: GPUs and Nvidia have already penetrated the supercomputer market and the entering was successful to say at least.
We hope for a revolution, but we always rely on evolution. What we have to understand is that any technology has a lot of ramifications. Those ramifications may result into further developments of products by the companies we never have heard of before. What we have to consider is that loads of technologies that actually change the world or mare it different come from big companies. Exceptions just prove the rules.